Commercial DNA Probes Reveal Colonization Patterns of Periodontopathogens in Chronic Periodontitis

RESEARCH REPORTS Commercial DNA Probes Reveal Colonization Patterns of Periodontopathogens in Chronic Periodontitis W. Keung Leung, Lijian Jin, Birgitta Söder, Esmonde F. Corbet In the present study, detection patterns of subgingival Porphyromonas gingivalis (Pg), Prevotella intermedia/nigrescens (Pi/n), Tannerella forsythia (Tf) and Treponema denticola (Td) were analysed, using commercially available species-specific DNA probes (DMDX/PathoTek, Switzerland) in 16 untreated periodontitis patients (5 7 non-adjacent sites per subject) before and one month after mechanical non-surgical periodontal therapy. The detection patterns of the pathogens at the two time points were studied using configural frequency analysis (CFA) and Stuart-Maxwell χ 2 test. All subjects showed favourable clinical treatment responses. Prevalence of all microbes followed was significantly reduced after treatment. CFA identified: (i) two statistically significant configural types at both time points: co-detection of Pg, Pi/n, Tf, Td, and none of the four pathogens detectable (Holm s criteria p = 0.003); and (ii) one statistically significant configural anti-type post-treatment: reduced detection pattern of Pi/n alone (Holm s criteria p = 0.004). For many of the pre-treatment sites with Tf, Pg, Pi/n, and Td configural type, all these periodontopathogens were undetectable one month after treatment (p < 0.001, McNemar s post hoc analysis with Bonferroni adjustment). Within the limitations of this study, the results indicated a co-detection of subgingival periodontopathogens Pg, Pi/n, Tf, and Td. Analysis of the subgingival colonisation of these microbes might be a useful adjunct in the assessment of periodontal therapeutic outcomes. Key words: Dental plaque/microbiology, periodontitis/microbiology, periodontitis/therapy, statistical model INTRODUCTION Chronic periodontitis is an infectious disease resulting from intricate host parasitic interactions at the soft-hard tissue border of the tooth-supporting apparatus (American Academy of Periodontology, 1996). The role of specific periodontopathogens in this disease process has been noted, dating back to the late 1970s (Socransky and Haffajee, 1994). Culture-based and molecular genetic techniques are frequently used to detect the fastidious micro-organisms that colonise the subgingival sites in humans (Zambon, 1996). Various commercial test systems or laboratories are in operation in different parts of the world for detection of periodontopathogens (Conrads, 2002). In view of the possible immense complexity of the microbial etiology of periodontal diseases, attention has been directed towards studying the interactions between members of the subgingival microbial biofilm (Simonson et al, 1992). Research evidence accumulated in recent years has pointed towards certain periodontopathogens tending to co-exist in subgingival niches of subjects with various perio- Perio 2006; Vol 3, Issue 2: 87 96 87

Table 1 Global K-th order log-linear analysis Marginals fitted Residual chi-squares Component chi-squares df L p df L p Baseline Pg, Pi/n, Tf, Td 11 173.490 0.000 Pg + Pi/n, Pg + Tf, Pg + Td, Pi/n + Tf, Pi/n + Td, Tf + Td Pg + Pi/n + Tf, Pg + Pi/n + Td, Pg + Tf + Td, Pi/n + Tf + Td 5 17.458 0.004 6 156.032 < 0.001 2 0.003 0.955 3 17.455 < 0.001 Pg + Pi/n + Tf + Td 0 0 1 2 0.003 > 0.999 One month post non-surgical therapy Pg, Pi/n, Tf, Td 11 149.571 0.000 Pg + Pi/n, Pg + Tf, Pg + Td, Pi/n + Tf, Pi/n + Td, Tf + Td Pg + Pi/n + Tf, Pg + Pi/n + Td, Pg + Tf + Td, Pi/n + Tf + Td 5 1.993 0.850 6 147.578 < 0.001 2 0.003 0.9593 3 1.990 < 0.75 Pg + Pi/n + Tf + Td 0 0 1 2 0.003 > 0.999 dontal diseases (Socransky et al, 1998; Jin et al, 1999). Various statistical analysis protocols have been applied to the processing of subgingival microbial data (Cohen, 1993; Ali et al, 1994; Socransky et al, 1998). Many studies employed statistic protocols focusing on two--two contingency, correlation or cluster analysis of target species. Configural frequency analysis (CFA) has been recommended as an appropriate statistical test, where configurations of the target micro-organisms in the subgingival environment (present, absent or in various combinations) could be potentially studied (Cohen, 1993). The main objective of CFA is to search for types and antitypes in the data. Types are the combinations of variables that exhibit frequencies (observed) that are statistically more frequent than what would be expected. Antitypes are the combinations of the variables that exhibit frequencies (observed) that are statistically less frequent than what would be expected. These types and antitypes draw investigators attention to exceptional microbial combinations. While cluster analysis is another commonly used statistical method for handling data of a large number of oral microbial variables (Socransky et al, 1998), the methodology has inherent limitations. The optimistic assumption, which is also the limitation of the method, is that clusters must be formed. Cluster analysis method ignores the possibility of, and cannot identify, disease-causing agents that may prefer not to form clusters with other agents, i.e. the anti-type(s). However anti-type(s) could be detected CFA. Furthermore, in cluster analysis, the statistical significance of the assumed clusters is not directly tested (Cohen, 1993). In more recent years, researchers have taken the monitoring of periodontopathogens forward investigating the concurrent clinical conditions and markers of host responses in periodontitis patients (Jin et al, 1999, 2002; Dar et al, 2001) and studying the relationship between prediction and assessment of periodontal therapeutic outcomes and the detection of combinations of periodontopathogens at various stages of periodontal therapy (Fujise et al, 2002). 88 Perio 2006; Vol 3, Issue 2: 87 96

Table 2 First-order configurations (model: Pg, Piln, Tf, Td) at baseline Configurations a Pg Pi/n Tf Td Observed frequency Expectede frequency I z I p Holm s Criteria p 19 1.30 0.0000 0.0031 type + 3 3.51 1.0643 b + 1 1.53 1.0924 b + 0 2.05 0.2510 b + 1 2.80 0.4519 b + + 0 4.14 0.0287 b + + 4 5.53 0.893 0.3720 + + 7 7.54 0.397 0.6916 + + 0 2.42 0.1717 b + + 0 3.30 0.0691 b + + 2 4.41 0.3529 b + + + 0 6.52 2.861 0.0042 > 0.0036 + + + 2 8.89 2.619 0.0088 + + + 3 11.87 2.776 0.0055 + + + 0 5.20 2.580 0.0099 + + + + 43 14.01 8.475 0.0000 0.0033 type a + = detectable, or = not detectable DNA probes b p-level computed using exact binomial The current study used CFA in studying subgingival microbial data obtained through the application of a commercially available DNA probe method. DNA probes have the benefits of targeting a limited range of species with reasonable sensitivity and specificity, and the commercial availability of this testing method facilitates its use in clinical practice. The microbes monitored were four frequently targeted periodontopathogens, namely Porphyromonas gingivalis (Pg), Prevotella intermedia/nigrescens (Pi/n), Tannerella forsythia (Tf), and Treponema denticola (Td). These microbes were studied in subgingival plaque samples of chronic periodontitis patients before and one month after non-surgical mechanical periodontal therapy. The null hypotheses of the current study were: (i) detection of each of the four target periodontopathogens in subgingival niches of chronic periodontitis patients would be independent of each other; and (ii) non-surgical mechanical periodontal therapy would not alter the detection frequencies and, in turn, the configurations of the four periodontopathogens studied. Perio 2006; Vol 3, Issue 2: 87 96 89

STUDY DESIGN AND RESULTS Subjects, Sampling and Examination of Samples Microbiological samples from 85 subgingival sites of 16 untreated Chinese adults (4 females, 11 smokers) aged 32 55 years (mean 43.4 ± 7.0 years) with chronic periodontitis were studied. Ethical Committee approval was granted for this treatment study, and no untreated control group was considered. Clinical and other findings (Jin et al, 2002) have been published. From each subject, 5 7 non-adjacent sites were sampled: 3 4 periodontitis sites [probing depth (PD) 5.0 mm, attachment loss (AL) 3.0 mm, radiographic crestal bone loss (BL) present], 1 2 gingivitis sites [gingival index (GI) 2, positive bleeding on probing (BOP), PD 4.0 mm, AL 1.0 mm, BL absent], and 1 clinically healthy site [GI 1, BOP absent, PD 3.0 mm, AL 1.0 mm, BL absent]. Samples were collected using a sterile, mediumsized paper point (Johnson and Johnson, New Brunswick, NJ, USA), which was gently inserted into the depth of the sampling site after drying the tooth and careful removal of all supragingival plaque. The paper point was left in place for 10 seconds and then was stored dry, in labelled sterile vials at -20 C. The microbiological samples were shipped chilled (4 C) in batches, within 4 weeks after sampling, for laboratory analysis (DMDx/PathoTek, ANAWA Laboratories, Zürich, Switzerland) for Pg, Pi/n, Tf and Td using four species-specific DNA probes. Tf was identified a recombinant DNA probe, while for Pg, Pi/n and Td genomic probes were used (Conrads, 2002). The Pi probe, as indicated in materials provided DMDx/PathoTek, cross-reacts with 'Type II Pi', i.e. the current Pn ATCC-33563-type strain. Therefore, positive detection this probe would be regarded as detection of Pi/n. The lowest detection levels of the commercial DNA probes used in the current study are 1000 cells/sample (French et al, 1986). Following sampling, each patient received a course of non-surgical periodontal therapy within a 6-week period as described previously (Jin et al, 2000); clinical details about the treatment responses in these patients appeared in a previous report (Jin et al, 2002). No antibiotics were prescribed and no patient reported, in response to questioning, antibiotic use for medical reasons over the treatment period. Subgingival plaque samples were collected from the same 85 sites one month after the completion of the non-surgical periodontal therapy, and these samples were handled and analysed the same methods. Statistical Analysis The microbial data were pooled. The computational unit was the sampling site. Stuart-Maxwell χ 2 test (Fleiss and Everitt, 1971) was used to analyse the change of microbial profiles, i.e. positive detection or not, of the four target species Pg, Pi/n, Tf and Td alone and in all possible combinations (n = 16, see Table 2) at baseline and one month post non-surgical therapy. Post hoc analysis based on McNemar s statistic with Bonferroni adjustment was carried out to test for significant microbial changes from baseline to one month posttreatment. The level of significance chosen was 0.05. CFA was then applied to the data set (Cohen, 1993). The protocol described von Eye (1990) was followed. A global K-th order log-linear analysis was performed to search for the appropriate statistical types. SPSS 11.0 for Windows software was used to calculate the expected frequency and component χ 2 p-value for the marginals fitted for models for CFA analysis. In brief, the first set of marginals fitted describes the model for mutual independence among all variables (the first-order model). The second, third and fourth models test, respectively, models based on two-, three-, and four-way interactions among the variables (Cohen, 1993). The statistical types were selected, based on the component of χ 2 p- value, and these were then put through the corresponding configuration analysis according to which types and/or anti-types were identified. The configuration analysis was conducted as follows. Firstly, the observed periodontopathogen configurations were put through a log linear regression analysis, based on the appropriate model detected, the model(s) with component of χ 2 p- value < 0.05. Then the expected frequency values for the different configurations were obtained. Based on that, the absolute z and then the p-values of the configurations were calculated. For situations where z could not be calculated, exact binomial approach using the observed and expected frequencies was employed to calculate the p-value (Cohen, 1993). The final p-values of the CFA were measured using Holm s Criteria (Cohen, 90 Perio 2006; Vol 3, Issue 2: 87 96

1993). A significant configural type was a configuration with a significant Holm s Criteria p with larger observed than expected frequency, while the contrary would be a significant anti-configural type, or anti-type. Clinical Parameters The clinical parameters of the subjects before and one month after the non-surgical periodontal therapy were as described earlier (Jin et al, 2002). In brief, the subjects had a mean of 26.0 ± 2.3 teeth, with 34.8 ± 18.9%, 44.9 ± 13.7% and 20.3 ± 13.1% healthy, gingivitis or periodontitis sites respectively. They showed favourable treatment responses to the single course of non-surgical periodontal therapy in terms of significant reductions of plaque levels (90.3 ± 13.9% to 58.5 ± 34.1%, p < 0.01), BOP (70.2 ± 20.1% to 44.0 ± 25.4%, p < 0.01), overall mean probing pocket depth (PPD) reduction (from 4.8 mm to 3.8 mm) and PPD reduction in sampling sites (from 5.7 mm to 4.5 mm, p < 0.01). Microbial Parameters The prevalence of Pg, Pi/n, Tf and Td was significantly (p < 0.001, Fisher exact test) reduced after treatment: 72.9%/35.3%, 55.3%/28.2%, 62.4%/17.6% and 68.2%/20.0% (baseline/ one month after treatment respectively). Study of the profiles of Pg, Pi/n, Tf and Td for individual subgingival sites at baseline and one month after non-surgical therapy revealed 11 out of the possible 16 patterns. In brief, patterns of detection of Tf alone and Pi/n + Tf, Pi/n + Td, Pg + Pi/n + Tf and Pi/n + Tf + Td co-detection were not observed at either time points. The overall Stuart- Maxwell χ 2 value of the 11 11 table (df = 10) was 46.77, which is considered significant, indicating a change due to treatment. When McNemar s post hoc analysis was carried out, with Bonferroni adjustment, overall sites with none of the four species detectable were significantly increased one month after non-surgical treatment, principally because most sites that at baseline had Pg, Pi/n, Tf and Td detected were found at one month after treatment to have none of these detectable (p < 0.001). The proportions of sites, grouped pre-treatment clinical condition, which had none of Pg, Pi/n, Tf and Td detected at baseline and at one month post-treatment were as follows: periodontitis sites, 3/54 to 26/54 (p < 0.001, Fisher exact test); gingivitis sites, 5/17 to 13/17 (p = 0.015, Fisher exact test); healthy sites, 11/14 to 13/14 (p = 0.6, Fisher exact test). The global K-th order log linear analysis of the baseline bacterial data showed that the first- and the second-order, independent or two-way interactions among the variables would be appropriate statistical types. While for the one-month post- nonsurgical therapy data, only the first-order, the independent model, would be the appropriate statistical type to be tested (Table 1). As shown in Table 2, two significant statistical microbial configural types were identified, co-detection of Pg, Pi/n, Tf and Td or none of Pg, Pi/n, Tf and Td being detectable. The same situation was observed upon first-order configuration analysis of data one month after non-surgical therapy (Table 3). From the same analysis, a significant statistical microbial configural anti-type of detection of Pi/n alone was observed. The results indicated preferential detection patterns of all of Pg, Pi/n, Tf and Td being detected or all these four species being undetectable. While non-surgical periodontal therapy resulted in many sites with Pg, Pi/n, Tf and Td at baseline becoming negative for detection the commercial DNA probe approach employed for all four species at one month, the two detection patterns of all four or none appear to be preferred both before and one month after therapy. However, at one month after non-surgical therapy, detection of Pi/n alone in subgingival plaque samples was not preferred. No more statistically significant configurations could be identified when second-order configuration analysis of data at baseline was incorporated into the model (results not shown). For comparison, the present set of data was also subjected to cluster analysis (Socransky et al, 1998), in which the similarity between pairs of the four species was computed using a φ coefficient, the coefficients were scaled and then sorted using an average unweighted linkage sort. In brief, both dendrograms of cluster analysis pre- and one month post-treatment showed all four species forming one cluster with similar levels of higher or equal to 80%. The clustering profiles were: (i) pre-treatment - both Pg and Td or Pi/n and Tf similarity of 86%, overall similarity of 85%; and (ii) one month post-treatment Pg and Pi/n similarity of 92%, Tf and Td similarity of 90%, overall similarity of 80%. Perio 2006; Vol 3, Issue 2: 87 96 91

Table 3 First-order configurations (model: Pg, Piln, Tf, Td) at one month post non-surgical therapy. Configurations a Pg Pi/n Tf Td Observed frequency Expectede frequency I z I p Holm s Criteria p 52 26.0 6.120 0.0000 0.0033 type + 5 14.18 2.671 0.0076 > 0.0038 + 1 10.23 3.077 0.0021 0.0036 antitype + 0 5.57 2.661 0.0078 + 1 6.50 2.449 0.0143 + + 9 5.58 1.279 0.2010 + + 1 3.04 0.8121 b,c + + 0 3.55 0.9734 b,c + + 0 2.19 0.8913 b,c + + 0 2.56 0.9257 b,c + + 1 1.39 0.8116 b + + + 0 1.20 0.7014 b,c + + + 2 1.40 0.3306 b + + + 1 0.76 0.3530 b + + + 0 0.55 0.0099 b + + + + 12 0.30 0.0000 b 0.0033 type a + = detectable, = not detectable b p-level computed using exact binomial DISCUSSION Genomic DNA probes (against Pg, Pi/n, Td) and recombinant DNA probes (against Actinobacillus actinomycetemcomitans, Campylobacter rectus, Eikenella corrodens, Fusobacterium nucleatum and Tf) have been widely used commercially to detect periodontopathogens in subgingival plaque in Europe, the USA and the rest of the world (e.g. Ali et al, 1994; Conrads, 2002; French et al, 1986; Jin et al, 1999, 2002; Porras et al, 2002). For instance, in 2000, about 15000 DNA-probe-based tests for periodontopathogens were performed in Germany and Switzerland (Conrads, 2002) for the guidance of periodontal treatment. Therefore, information on the use and interpretation of results of this widely used test method would be helpful. The present report focused on a clinical approach of utilising the commercially available and widely used DNA probes for detection of periodontopathogens to explore how data generated from these tests may be used with respect to bacterial combinations to assist in the clinical management of various forms of periodontitis. The lowest detection levels for the various DNA probes used in this study were similar to those of the DNA-DNA checkerboard hybridisation technique (Socransky et al, 1994) and the quantitative real-time PCR technique (Asai et al, 2002). The 92 Perio 2006; Vol 3, Issue 2: 87 96

DNA-DNA checkerboard approach can, however, distinguish between Pi and Pn (Socransky and Haffajee, 1994). Within healthy American subjects, the mean prevalence and levels of Pi and Pn at subgingival sites were similar. Pi and Pn have greater, but similarly equal, prevalence at higher quantities in American periodontally diseased subjects (Haffajee et al, 1998). However, a highly specific PCR study on 97 samples from 31 Scottish periodontitis subjects showed a similar prevalence to periodontitis American subjects of Pi at 40% but not of Pn (Riggio et al, 1998). It seems as if the preferred 'gold standard' detection method for Pn and this microbe s role in periodontitis are issues yet to be clarified. The present study analysed the detection of Pi and/or Pn as a group, i.e. Pi/n. A recent review concluded that endodontic paper point sampling may under-represent the microbes at the base of a periodontal pocket, while curette sampling would better represent the bacteriology of the entire pocket (Loomer, 2004). The same article also drew attention to the variability in microbiology among different pockets of the same individual, which remains an unresolved issue in microbial monitoring of periodontitis patients, unless full-mouth sampling is conducted. Multiple samples from 3 4 pockets were collected in the present study. Higher proportions of putative pathogenic bacteria have been reported to be detected using paper point sampling compared with curette sampling (Renvert et al, 1992). Thus the paper point sampling technique employed should maximise the chances of the detection of the four target species. Van der Velden and co-workers (2003) reported that smokers and non-smokers with chronic periodontitis had similar prevalence of 7 periodontopathogens, while one similar study reported, among the 5 periodontopathogens followed, higher pre-treatment prevalence of Tf and Td in smokers (Dar et al, 2005). Both studies reported that in smokers the prevalence of Tf remained significantly higher post-therapy with an increased prevalence of Pi in smokers post-treatment (van der Velden et al, 2003). The fact that 11/16 (69%) of the study cohort were smokers might imply that Tf and perhaps Pi might be more readily detectable one month after non-surgical therapy. The results of the present study should therefore be interpreted with full awareness of its limitations and the composition of its subjects. Cohen (1993) employed CFA method in re-analysis of published subgingival microbial data in culture (Slots et al, 1986) and culture-independent (Simonson et al, 1992) studies. The anaerobic culture study showed that the presence of Pg (Bg, previously Bacteroides gingivalis) alone was preferred in samples from subjects with untreated- or treated-progressive periodontal disease, whereas the absence of A. actinomycetemcomitans, Pg and Pi (Bi, previously Bacteroides intermedius) configuration was not preferred (Cohen, 1993). On the contrary, absence of A. actinomycetemcomitans, Pg and Pi was preferred in samples from patients with no-disease or treated-non-progressive periodontal disease. The culture (Slots et al, 1986), ELISA (Simonson et al, 1992) and DNA probe (present study) approaches in periodontal microbiological testing, as discussed Aass et al (1994), all probably do not share the same detection levels and each would have evidenced its own inherent limitations for the detection of microbes studied. A preferred combination of Pg and Tf was demonstrated use of monoclonal antibodies and a very sensitive indirect immunofluorescence technique (Gmür et al, 1989). Socransky and coworkers (1998) published a study using the checkerboard DNA-DNA hybridization detection system for detection of bacteria in microbial samples collected from every mesiobuccal site in healthy subjects and periodontitis patients. A total of 13261 samples were processed and cluster analysis was employed. CFA was used in a focused fashion only to confirm the data from individual elements of the cluster analysis. In theory, CFA analysis should preferably be carried out before the cluster analysis. Nevertheless, a preferred configuration of Pg, Tf (Bf, previously Bacteriodes forsythus) and Td, the so called red complex, in subjects with periodontal disease was reported. The relatedness of these three species to Pi or Pn, as shown the same study, was slightly higher than 40%. The red complex has been extensively reviewed (Holt and Ebersole, 2005), but the codetection of only the three periodontopathogens of this complex may not be the only predominant detection pattern. The relatedness of the four periodontopathogens identified in the current group of patients pre- and post-therapy, as shown CFA or cluster analysis, was high. Pi/n appeared to relate closer to Pg, Tf and Td in plaque samples of Perio 2006; Vol 3, Issue 2: 87 96 93

this group of predominantly male and smoking Chinese subjects than in the American cohort reported Socransky et al (1998). What could not be observed cluster analysis alone was the configural type of the non-detectability of Pg, Pi/n, Tf and Td at both time points and the non-preferred detection of Pi/n alone one-month post non-surgical periodontal therapy. The present study dealt with both pre- and posttherapy microbial detection profiles of four putative periodontopathogens in an attempt to understand the nature of interactions among these pathogens as members of the subgingival biofilm before and following mechanical periodontal treatment, along with the changes in local environment consequent to therapy, such as reductions in pocket depths. Dar and co-workers (2001) showed that in 33 untreated adult periodontitis patients, scaling and root planing could produce statistically significant reductions of the prevalence of Pi, Tf and Td, detected the PCR method, in 132 periodontitis sites studied. The frequency of Td post-therapy was reduced but not statistically significantly. The present study showed that mechanical therapy was successful in reducing a predominant co-detection of the four species, i.e. Pg, Pi/n, Tf and Td into a predominant nil detection of the four species. Our earlier findings suggest that persistent co-detection of the four species one month post-treatment is associated with poorer treatment outcomes (Jin et al, 2002). However, it is yet to be demonstrated if this association reflects a cause and effect relationship. Taken together, previous studies and the present study suggest that certain periodontopathogens, particularly Pg, Pi/n, Tf and Td, if detectable in subgingival plaque samples from inflamed sites using commercially available DNA probes, tend to be co-detected, which was apparently independent of the presence of periodontal pockets, thus not substantiating the first null hypothesis of this study that the pathogens were detectable independent of each other. Interestingly, undetectable levels of pathogens were also found in untreated (and possibly non-progressive) diseased sites. The reasons for non-detection of the target pathogens might be either true absence or, more likely, presence at quantities below the detection limits, given as 1000. The non-surgical mechanical periodontal therapy seemed to produce an effect of rendering the pathogens undetectable, implying that the nonsurgical periodontal therapy could alter the detection frequencies of the four periodontopathogens. No non-treated group as a control was included, but the microbial shifts are unlikely to occur other than in response to treatment. Longitudinal microbial and clinical monitoring of diseased sites seemingly negative for the target pathogens may elucidate microbial associations and reflect the sporadic nature of periodontitis (Goodson et al, 1982). Ali and co-workers (1994), in their study using the same bacterial detection protocol, had shown that Pg, Pi/n, F. nucleatum and Tf were positively correlated in periodontally untreated Sudanese with adult chronic periodontitis. Lotufo and co-workers (1994), using a different DNA probe detection method, showed Pg occurrence at subgingival sites to be associated with Tf. A study using 16 S rrna gene-detected multiplex PCR protocol, reported the simultaneous detection of Pi and Tf in various sites, including 38 subgingival plaque samples from untreated periodontitis patients (Conrads et al, 1999). Both species were present in 13% of the subgingival specimens, while 55% of the specimens were negative for Pi and Tf, but no statistical tests were performed to examine the association between the two bacteria. The present study revealed the co-detection of Pg, Pi/n, Tf and Td using commercially available DNA-probe analysis, further confirming the previous observations suggestive of co-existence of certain periodontopathogens, possibly because of nutritional co-dependencies or adhesion (Wyss, 1992). The CFA method of analysing periodontopathogen-detection patterns, despite its potential to be used to promote understanding and interpretation of otherwise quite complicated microbial data, is not without limitations. The pooled microbial data of the present study was, however, sufficiently powerful to allow for analysis. Commercial periodontopathogen-detection services could potentially generate databases that could, when studied, improve understanding of the preferred co-inhabitant patterns of periodontopathogens. Knowledge of the co-inhabitant patterns of periodontopathogens in periodontally diseased and periodontally healthy conditions could be of use to clinicians in their daily practice in the treatment and especially in the delivery of supportive periodontal care of chronic periodontitis, if they use diagnostic commercial DNA periodontopathogendetection services. 94 Perio 2006; Vol 3, Issue 2: 87 96

In conclusion, using a commercial DNA-probebased detection method, the periodontopathogens Pg, Pi/n, Tf and Td were shown to exhibit a tendency to be co-detected in periodontal sites all four together or not at all. Non-surgical periodontal therapy was shown to change a large proportion of sites from being positive for Pg, Pi/n, Tf and Td to all Pg, Pi/n, Tf and Td being below detectable levels. However, in certain sites the co-detection of the four species persisted. The implications of these observations in the context of the use of commercially available DNA-probe services for detection of periodontopathogens and how results from such services may usefully inform periodontal treatment require further investigation, if possible correlating the detection patterns at sites with clinical and/or host risk or biomarker data at the same sites. Acknowledgements The work described in this paper was partially supported grants from the Research Grants Council of the Hong Kong Special Administrative Region, China (HKU 7310/00M to LJ and HKU 7331/00M to WKL). REFERENCES Aass AM, Preus HR, Zambon JJ, Gjermo P. Microbiologic tests in epidemiologic studies: are they reproducible? Scand J Dent Res 1994;102:355 360. Ali RW, Skaug N, Nilsen R, Bakken V. Microbial associations of four putative periodontal pathogens in Sudanese adult periodontitis patients determined DNA-probe analysis. J Periodontol 1994;65:1053 1057. American Academy of Periodontology. Consensus Report. Periodontal Diseases: pathogenesis and microbial factors. Ann Periodontol 1996;1:926 932. Asai Y, Jinno T, Igarashi H, Ohyama Y, Ogawa T. Detection and quantification of oral treponemes in subgingival plaque real-time PCR. J Clin Microbiol 2002;40: 3334 3340. Cohen ME. Configural frequency analysis for exploring bacterial sets in periodontal health and disease. Community Dent Oral Epidemiol 1993;21:283 287. Conrads G. DNA probes and primers in dental practice. Clin Infect Disease 2002;35(Suppl 1):S72 S77. Conrads G, Flemmig Tf, Seyfarth I, Lampert F, Lütticken R. Simultaneous detection of Bacteroides forsythus and Prevotella intermedia 16 S rrna gene-directed multiplex PCR. J Clin Microbiol 1999;37:1621 1624. Dar IB, Hodge PJ, Riggio MP, Kinane DF. Clinical and microbiological effect of scaling and root planing in smoker and non-smoker chronic and aggressive periodontitis patients. J Clin Periodontol 2005;32:200 206. Dar IB, Mooney J, Kinane DF. Changes in subgingival microflora and humoral immune response following periodontal therapy. J Clin Periodontol 2001;28: 796 805. Fleiss JL, Everitt BS. Comparing marginal totals of square contingency tables. Br J Math Statist Psychol 1971;24: 117 123. French CK, Savitt ED, Simon SL, Eklund SM, Chen, MC, Klotz LC et al. DNA probe detection of periodontal pathogens. Oral Microbiol Immunol 1986;1:58 62. Fujise O, Hamachi T, Inoue K, Miura M, Maeda K. Microbiological markers for prediction and assessment of treatment outcome following non-surgical periodontal therapy. J Periodontol 2002;73:1253 1259. Goodson JM, Tanner ACR, Haffajee AD, Sornberger GC, Socransky SS. Patterns of progression and regression of advanced destructive periodontal disease. J Clin Periodontol 1982;9:472 481. Gmür R, Strub JR, Guggenheim B. Prevalance of Bacteroides forsythus and Bacteroides gingivalis in subgingival plaque of prosthodontically treated patients on short recall. J Periodontal Res 1989;24:113 120. Haffajee AD, Cugini MA, Tanner A, Pollack RP, Smith C, Kent RL Jr, et al. Subgingival mircobiota in healthy, wellmaintained elder and periodontitis subjects. J Clin Periodontol 1998;25:346 353. Holt SC, Ebersole JL. Porphyromonas gingivalis, Treponema denticola, and Tannerella forsythia: the red complex, a prototype polybacterial pathogenic consortium in periodontitis. Periodontol 2000 2005;38:72 122. Jin LJ, Leung WK, Corbet EF, Söder B. Relationship of changes in interleukin-8 levels and granulocyte elastase activity in gingival crevicular fluid to subgingival periodontopathogens following non-surgical periodontal therapy in subjects with chronic periodontitis. J Clin Periodontol 2002;29:604 614. Jin LJ, Söder P-Ö, Leung WK, Corbet EF, Samaranayake LP, Söder B et al. Granulocyte elastase activity and PGE2 levels in gingival crevicular fluid in relation to the presence of subgingival periodontopathogens in subjects with untreated adult periodontitis. J Clin Periodontol 1999;26:531 540. Jin LJ, Wong KYN, Leung WK, Corbet EF. Comparison of treatment response patterns following scaling and root planing in smokers and non-smokers with untreated adult periodontitis. J Clin Dent 2000;11:35 41. Loomer PM. Microbiological diagnostic testing in the treatment of periodontal diseases. Periodontol 2000 2004; 34:49 56. Lotufo RF, Flynn J, Chen C, Slots J. Molecular detection of Bacteriodes forsythus in human periodontitis. Oral Microbiol Immunol 1994;9:154 160. Porras R, Anderson GB, Caffesse R, Narendran S, Trejo PM. Clinical response to 2 different therapeutic regimens to treat peri-implant mucositis. J Periodontol 2002;73: 1118 1125. Renvert S, Wikström M, Helmersson M, Dahlén G, Claffey N. Comparative study of subgingival microbiological sampling techniques. J Periodontol 1992;63:797 801. Riggio MP, Lennon A, Roy KM. Detection of Prevotella intermedia in subgingival plaque of adult periodontitis patients polymerase chain reaction. J Periodont Res 1998; 33:369 376. Perio 2006; Vol 3, Issue 2: 87 96 95

Simonson LG, McMahon KT, Childers DW, Morton HE: Bacterial synergy of Treponema denticola and Porphyromonas gingivalis in a multinational population. Oral Microbiol Immunol 1992;7:111 112. Slots J, Bragd L, Wikström M, Dahlén G. The occurrence of Actinobacillus actinomycetemcomitans, Bacteroides gingivalis and Bacteriodes intermedius in destructive periodontal disease in adults. J Clin Periodontal 1986; 13:570 577. Socransky SS, Haffajee AD. Evidence of bacterial etiology: a historical perspective. Periodontol 2000 1994; 5:7 25. Socransky SS, Haffajee AD, Cugini MA, Smith C, Kent RL Jr. Microbial complexes in subgingival plaque. J Clin Periodontol 1998;25:134 144. Socransky SS, Smith C, Martin L, Paster BJ, Dewhirst FE, Levin AE. Checkerboard DNA-DNA hybridization. Biotechniques 1994;17:788 792. van der Velden U, Varoufaki A, Hutter JW, Xu L, Timmerman MF, van Winkelhoff AJ et al. Effect of smoking and periodontal treatment on the subgingival microflora: a retrospective study. J Clin Periodontol 2003;30:603 610. von Eye A. Introduction to configural frequency analysis: the search for types and antitypes in cross-classifications. Cambridge: Cambridge University Press 1990;16 26. Wyss C. Growth of Porphyromonas gingivalis, Treponema denticola, T. pectinovorum, T. socranskii, and T. vincentii in a chemically defined medium. J Clin Microbiol 1992;30:2225 2229. Zambon JJ. Periodontal diseases: microbial factors. Ann Periodontol 1996;1:879 925. Reprint requests: W. K. Leung Periodontology, Faculty of Dentistry The University of Hong Kong Room 3B39, 34 Hospital Road, Hong Kong SAR, China. Tel: +852 2859 0417 Fax: +852 2858 7874 Email: ewkleung@hkucc.hku.hk 96 Perio 2006; Vol 3, Issue 2: 87 96